This text of mine is a part of my continuation studies. I wrote it in the begining of the 1990’s. It deals with universalism and relativism in psychology, particularly with linguistic universality and relativity.
The former argues that the cognitive base of all languages is the same while the latter holds that thinking may be different depending on the language of the thinker. The evidence for both sides is critically reviewed and both are found to be lacking in some respects. It is claimed that proponents of both sides may be in error in concentrating their research on linguistic phenomena – especially words, sentences and their meaning while ignoring the psychological level.
Syntax learning is considered to be of minor importance in the study of language learning. Cognitive science does not seem to be interested in comparative psychological linguistics. The powerful phenomena of the image have been almost totally ignored. A theory closing the gap between the image and syntax is reviewed in detail and is found to settle many questions.
INTRODUCTION
The question of universalism and relativism in psychology is an old one. The ancient Greeks were aware of cultural differences. As the famous maxim of Protagoras put it: “When in Athens do as the Athenians do.” Sokrates, however, managed to obscure for future generations the possibility that relativism might be aligned with realism – that there may be spatiotemporally indexed “facts of the matter” (Fuller, 1991).
The striving in psychology (as well as in other sciences) is for universal truths. Most of the psychological research is considered to be valid for all people whatever their cultural or linguistic background might be. Processes of e.g. learning, memory and group formation seem to hold for all nations. Therefore it quite as easy to believe in cognitive universalism as it is to believe in cognitive relativism. Both sides have their proponents as well as their opponents and both sides may cite research in favour of their opinion.
The best-known proponent of the universalist side is Noam Chomsky, a linguist from whose writings it can be inferred that his theory of language is also a psychological one (Herriot, 1970). The other side relies also heavily on linguists, namely on Edward Sapir and B. Lee Whorf. Chomsky’s theory is expressed in terms of the generative grammar. The connections between language and thought expressed by Sapir and Whorf are known as the Sapir-Whorf hypothesis or the linguistic relativity hypothesis.
The question of universalism and relativism is discussed mainly in relation to language, so this paper deals mainly with language universality and language relativity. The first part discusses in brief varieties of relativism. The second chapter is devoted to the ideas of Sapir and Whorf. The third part examines the theories put forth by Chomsky and Fodor. The final chapter presents a semiotic theory of imagery processes proposed by Frode J. Strømnes.
1. Varieties of relativism
Kamppinen and Revonsuo (1991) list several varieties of relativism. According to them there exist six basic kinds of relativism. They are: descriptive and normative cognitive relativism, descriptive and normative ethical relativism, ontological relativism and axiological relativism. As the seventh they name cultural relativism, which is a mixture of normative and descriptive relativism.
Descriptive cognitive relativism consists of the factual claim that there are apparently incompatible beliefs about the world. It is an empirical fact. In anthropology it is the starting point of research (ibid, p. 36). Normative cognitive relativism claims that there are no reasonable or definitive grounds for deciding between different beliefs and therefore we should not pass a judgement concerning the apparently incompatible beliefs.
The weak version of cognitive relativism states that there are incompatible beliefs only, but the grounds and rules of reasoning are universal. As such it should not be considered relativism at all, but as cognitive universalism. According to the strong version the rules of reason and even the meta-standards of rationality can be different in distinct cultures.
Ontological relativism is the metaphysical conclusion that can be drawn from the fact of descriptive cognitive relativism. It holds that the apparently incompatible beliefs are grounded upon a multitude of incompatible realities. In other words, the incompatible beliefs are about different worlds. Axiological relativism is the metaphysical conclusion drawn from descriptive ethical relativism. It claims that there are multiple, mutually incompatible values out there.
Kamppinen and Revonsuo criticize (1991) cultural relativism on conceptual grounds, but there are also more practical reasons for cultural relativism to be a source of confusion. As moral norms, habits and language at least seem go hand in hand and vary together between cultures, it is not always easy to distinguish between them when studying groups with varying cultural backgrounds.
The classification of Kamppinen and Revonsuo is only one of many possible. Niiniluoto (1991) classifies two main types of relativism, viz. cognitive and moral. We can also separate subtypes within cognitive relativism such as perceptual relativism (Krogh, 1991).
2. Linguistic relativity
The most ardent proponents of linguistic relativity were Benjamin Lee Whorf and his teacher Edward Sapir. Devitt and Sterelny (1987) summarize their position as follows:
1. All thinking is “in a language – in English, in Sanskrit, in Chinese”. (Whorf, 1956: 252)
2. Each language structures a view of reality.
3. The views of reality structured by languages, or at least by families of languages, differ.
There are two versions of the linguistic relativity principle. The strong form – linguistic determinism – states that cognition and thought processes are totally determined by the structure of the language that we speak. The weak form – linguistic relativity – says that the structures of different languages do exert some influence on the thinking and categorization of speakers (Penn, 1972: 28-32).
Neither Sapir nor Whorf was very clear about that question. The stronger interpretation is expressed in a passage like: “We cut nature up, organize it into concepts, and ascribe significances as we do, largely because we are parties to an agreement to organize it in this way – an agreement that holds throughout our speech community” (Whorf, 1956, 213-214). At other times he seems to be in favour of the weaker interpretation: “Newtonian space, time, and matter are no intuitions. They are receipts from culture and language. That is where Newton got them (ibid, p.153). Sapir wrote: “We see and hear and otherwise experience very largely as we do because the language habits of our community predispose certain choices for interpretation (1929, p. 210). No two languages are ever sufficiently similar to be considered as representing the same social reality” (1949, p. 162).
Later proponents of the Sapir-Whorf hypothesis (at least most of them) seem to believe in the weaker version of the theory: “The admitted possibility of translation between languages of diverse structures spoken by people of different cultures is scarcely compatible with total linguistic determinism, while the equally admitted difficulties involved in translation afford solid support for the validity of linguistic relativity” (Robins, 1976).
Fuller (1991) takes the pro-Whorfian side in a more philosophical debate over relativism: “For, if the mere existence of one world were sufficient to cause different people to experience a world that they presume others also to experience, then there should be no diversity at all. However, the fact that diversity exists suggests that cultures unwittingly presume different worlds of one another”.
The research into the question of linguistic relativity has concentrated on vocabulary and syntax. This is because thoughts are considered to be linguistic or at least to come very close to being it. Some writers take a definite stand:” What kind of inner representations are thoughts? At the very least, they must be language-like in character” (Devitt & Sterelny, 1987, p. 116; my italics). “Images, maps and diagrams may be associated with thoughts, particularly with perceptually based ones, but they are not themselves thoughts. In sum, many thoughts are unpicturable; any picture could be associated with many thoughts. Thought and talk are abstract, and abstract in the same way. It seems clear that our system of inner representation and our system of linguistic representation are importantly similar” (p. 117; italics in original).
Some others do not know what to think. Herriot (1970), when discussing Chomsky’s theory, writes: “The assumption is the basically incorrect one that a sentence is the unit of meaning, although this is preferable to the assumption that the word is the basic unit.”
Whorf was a linguist. He studied the life and language of Hopis, a tribe living in the USA. He spent many years with them and knew what they meant with certain expressions. His knowledge, of not linguistics alone, but also of Hopi life led him to his hypothesis. It may be assumed that the linguistic means that he possessed were not satisfactory instruments to express the psychological meaning of the grammatical categories of Hopi. This fact had to be admitted even by Helmut Gipper, whose linguistic analysis of the Hopi conception of time led him to conclude that it cannot differ radically from that of SAE (Standard Average European), but “in reality the problem is much more complicated” (Gipper, 1976). In spite of his linguistic analyses he had to admit that “… it is not our time; it is rather the duration of certain ritual events relevant to Hopi life”.
3. Chomsky’s transformational grammar
Chomsky formulated his transformational generative grammar or as it is most often called transformational grammar as early as 1957 and revised it 1965. His theory is not an empirist one but he derives linguistic descriptions from his own intuitive analysis of language (Herriot, 1970). However, his theory of language has produced a lot of psychological research, e.g. that of Miller (1962) and many others after him.
Chomsky’s theory was revolutionary in that it distinguished between surface structure that is represented by the words that are spoken or written and deep structure, which is the underlying, more abstract (?) meaning of the sentence. He pointed out that two sentences may have very different surface structures but very similar deep structures and vice versa. Consider the following examples (Matlin, 1989): Sara threw the ball/The ball was thrown by Sara and John is easy to please/John is easy to please. The former sentences have very different surface structures but their meaning is the same. The latter have similar surface structures but very different deep structures.
The most interesting part of Chomsky’s theory, and at the same time the most controversial, is that of innateness. He claims that there are grammatical rules common to all languages – “universal grammar” – which are innately known by speakers (Chomsky, 1972). Not only does he claim that there is universal syntax but that there is also universal phonetics and universal semantics. Chomsky put forward three theses (Devitt and Sterelny, 1987):
1. Human beings are innately predisposed to learn languages.
2. Humans have an innate, richly structured, language specific, learning device.
3. The innate language-acquisition device consists in propositional knowledge of universal grammar.
It is easy to believe in the first one. Humans learn languages, animals do not. The second one is more interesting because it conflicts with the view that the innate structures that make language learning possible are the ones that make all learning possible. In his third hypothesis he goes against any empirism, traditional or contemporary (Devitt and Sterelny, 1987). Why could not a general learning device do the same thing? The problem is that to this day no one has succeeded in construing such a general learning device.
The reason why Chomsky goes this far is that he considers the grammatical rules picked up by the language learner to be abstract, subtle and unobvious (e.g. Chomsky, 1972). Here Chomsky considers language as something very distant from everyday life.
He studied language in isolation. Chomsky is a rationalist, not an empirist. If he would consider language as something concrete, something that refers to observable relations in the world, he would not have to presume that there exists an innate language learning device. This position will be examined in detail in the last chapter.
Chomsky is not alone with his innateness thesis. Similar thoughts have been put forth by Jerry Fodor. He thinks that there exists an innate Mentalese with the conceptual resources of any public language: “… at least one of the languages which one knows without learning is as powerful as any language that one can ever learn” (Fodor, 1975, p.82).
Fodor’s theory parallels that of current learning theory. According to him learning consists literally in the formulation and testing of hypotheses. These hypotheses must be framed in a system of representation. As to concept learning Fodor thinks that concepts can only be learnt by the construction of internal representations that define that concept. Learning the concept video camera is possible only if the learner already has the resources to define that concept. Undefinable concepts cannot be learnt at all (Fodor, 1981).
It is easy to agree with Fodor this far, but when it comes to language learning he makes two serious mistakes. First, he confuses the meaning of the word with its acoustic (or written) expression. Second, his line of thought would lead us to assume that animals have innate concepts. According to Fodor (1975) when learning the word tiger the learner must internally represent a hypothesis along the following lines: ‘tiger’ applies to x if and only if x is a tiger. Thus to learn the word for tigers, one must already have a way of representing tigers; one must have the concept of tiger or an equivalent complex of concepts.
If the acoustic (or written) form of a word would have meaning, we would have only one language in the whole universe (and the whole problem of language relativity would be non-existent). Words are only arbitrary signs agreed upon within the language community (de Saussure, 1966) . So we can have different words for a girl in different languages: Mädchen in German and flicka in Swedish. The meaning of words has to be sought somewhere deeper.
If learning words presupposes internal concepts, we would have to assume that parrots have a concept for tigers. We know that parrots can learn words. Reasoning along these lines would lead one to assume that parrots (that have never seen tigers) have innate concepts for them. This would lead us to credit parrots with innate Mentalese and capability to learn languages. This would be stretching the imagination too far.
Devitt and Sterelny (1987) argue in favour of an innate language-acquisition device on the following grounds:
1. It is done very young.
2. The data the child receives is degraded, often not consisting of good sentences.
3. There is little explicit instruction.
4. The level of acquisition is quite uniform by comparison to other intellectual skills.
5. All children, whatever their language, acquire elements of linguistic capacity in the same order.
Let us examine these arguments in detail. For the first one, many other skills are learnt very young. The child learns to walk (a very complex motor skill) much earlier than he or she learns to talk. By the time the first words appear the child can do many things that are at least as complex as speech. So there is nothing special in language learning in this sense.
Devitt and Sterelny argue that the data the child receives is degraded. It is degraded in a certain way. It is just because the child hears simplified utterances that he or she can extract meanings. No one teaches the beginning violinist the masterpieces of Paganini as the very first ones. First we have to know the elements of language before we can compose more complicated sentences. Language learning does not succeed by having children exposed to language as we know of children of deaf parents. Watching TV and hearing people utter good sentences does not help. The child does not learn language.
As to the amount of explicit instruction there is plenty of it. The mother usually speaks a lot with the child. All of the family members are eager to hear the child utter the first words and as the child does that he or she is helped all the way to sharpen those linguistic skills. If this is not explicit instruction then what is. It does not have to take place in the classroom to be successful. In a classroom it would not be that successful.
The elements of language are so simple that it may seem that all children acquire language at the same pace. This is not the situation at the age of six, however. There are observable differences in the linguistic skills of children of this age. The children of more educated parents (usually) have a better command of their mother tongue. They have been exposed to more varied and complicated language than children of less educated parents. The same goes for other skills. Think of children in the circus. They have command of motor skills most children would even not dream of. Language is just a skill than can be practiced as any other skill.
The last argument says that all children acquire elements of linguistic capacity in the same order. There is research evidence that Finnish and English children do not acquire these elements in the same order. This might well be connected to the hypothesis put forth by Strømnes (1974) that the verbs and substantives in these two languages do not carry the same meaning psychologically.
Though Devitt and Sterelny argue for an innate language-leaming device they, too, reject the view that speakers have innate knowledge of universal grammar or of anything else about their language. It is opinion of the present author that we do not have to assume a language learning device. I will now turn to consider the psychological theory of language by Frode J. Strømnes in contrast to the very linguistic one by Noam Chomsky.
4. A semiotic theory of imagery processes
The theory of Strømnes (2006) is clearly on the relativist side but it goes one step deeper than Chomsky and Whorf in the explanation of differences in the cognition of different languages. This is the level of the central nervous system.
We discussed meaning in the previous chapter. Meaning, in the opinion of Strømnes, lies in the images the words are capable to bring about. Images are inner representations analogous to vision. Vision is therefore the primary sense in language learning.
When we learn words we learn acoustic (or written) representations for our images. According to his model we may have several storage spaces for semantic representations, one for each language we know. The words girl, flicka and Mädchen all make us to ‘see’ an image of a girl provided that we know the respective languages. So we can have several semantic storage spaces but only one space for meaning, the analogue representation of images (Strømnes, 1974a).
This simple assumption explains why we can learn words, even in several languages. We see an object and hear sound waves that are linked to the appearance of that object. That is how it begins: the mother utters the word “mama” several times when taking care of the child. Eventually the child learns that this combination of sound waves refers to mother. Words are primarily learned by associating them with what is perceived (Strømnes and Iivonen, 1985). First there are single word utterances, then combinations of two words and so on. The child gets feedback all the time. The meaning of words is sharpened little by little until it is correct. But there is more to language and language learning than that. Language is a means of communication and communication, according to Strømnes means transferring combinations of images from the sender to the receiver. How is this done?
Combinations of images are entities too large to go into your nervous system through the ear all at one time. They have to be transformed into strings (acoustic or written strings; this time it is in written words). Think of a vase and a table. The vase stands on the table. Now you have a combination of two images with a certain geometric relation between them. How did that geometric relation come about? The critical word is the preposition ‘on’. It tells you to combine the images of the vase and the table in a certain way. It is ‘vase on the table’ not ‘table on the vase’. So the prepositions in the SAE-languages are the words that tell you how to combine images to get the message. In the Finnish language (and in many more, e.g. Estonian, Hungarian and Japanese) the same thing is done by suffixing words with case-endings. (Strømnes, 1973)
If the prepositions and case-endings would translate one to one, there would be nothing to be relativistic about – but they do not. It is not only that they do not translate one to one, but they form different geometric systems. This was demonstrated by Strømnes (1973) when he devised animations of the Swedish prepositions and the Finnish cases. The experimentally found system of Swedish (and probably the other Indo-European languages) is a vector geometric system. The meaning of the preposition in the vector system is carried by movement in three-dimensional space. So we have: ‘up’, ‘down’, ‘through’ and ‘around’ for example. Now we see what Whorf meant with his remark on Newton. Vector geometry is a product of language – the Indo-European.
The system of Finnish is a topology. The meaning in the case-endings is carried by the relations of the borders of two figures. We have e.g. ‘-ssa’ where the borders of the first figure are totally inside the second one. There is much less need of movement in this system. Actually the text attached to the films needed no verb. As all native speakers of Finnish know, the two-substantive expression “Mattiko papiksi?” (Is Matti going to become a priest?) is perfectly understandable. The substantive in the Finnish language carries some meaning which, in the Indo-European languages, only goes with the verb. The texts for the Swedish films needed a verb to be understandable. In the experiments nonsense words were used. The Swedish film had a substantive, a verb (in past tense) and the preposition. In the Finnish films a combination of two substantives were used. The case ending was attached to the second one.
Studies on elliptic sentences (newspaper headings and commentaries of ice-hockey matches) produced results in favour of the hypothesis about the significance of the verb and the substantive. The Finnish newspaper headings and the Finnish commentaries contained more sentences without a verb than the Swedish ones. A sentence without a verb was also judged as more grammatical in Finnish by Finnish subjects than was the same sentence in Swedish by Swedish subjects (Strømnes, 1974b). Finnish ice-hockey reporters spoke also more about the groupings of the players than their Swedish-speaking colleagues, who in turn spoke more about the movement of the puck. The Finns were thus more interested in the substantive-kind of information than the Swedish-speaking journalists or the ice-hockey reporters, in whose interest it was to use more verbs.
The time systems of these languages differ, too. Each case-ending seemed to have duration of its own designating different time. This was not very well studied at the time when the films were made (Strømnes, 1974c). Later research on classical plays filmed in the Nordic countries (Strømnes, Johansson and Hiltunen, 1982) demonstrated that the Swedes and Norwegians took great care to keep track of continuous time in the films, whereas the Finns concentrated on the relations of the characters.
The Swedes and the Norwegians were also very careful in keeping the viewer aware of the space of the play. When watching the films one was always able to tell where the actors came from and where they went. This was not the case with the Finnish versions. The narrative in the Finnish films was on a much deeper emotional level than in the Swedish or Norwegian ones. The overall differences between the films were so great that it was sometimes difficult to believe that they were based on the same manuscript.
The above holds for related languages, too. We recorded and compared some Finnish and Estonian materials. The Estonian material displays the same pictorial structures as the Finnish. We also had the opportunity to compare a Hungarian and British rendering of the same manuskript. The film research will be discussed in detail in later posts.
These studies strongly contradict the slogans ‘picture language is universal’ or ‘television language is universal’. Finnish late 1900-films used to sell badly in Europe, not because that their quality was bad, but because the pictures have a logic different from the Indo-European way of rendering. The best-selling Finnish films in the 1970’s were made by Swedish-speaking directors living in Finland (Ihamuotila, 1979). Nowadays Finnish movies by Finnis-speaking directors have had some success, at least in Europe.
Strømnes’ conception of language proved to be effective in teaching the deaf. Prelingually deaf children with nonverbal IQ at or below the normal level constitute a hopeless group of syntax learners. Strømnes and Iivonen (1985) managed, however, to teach a group of boys with a nonverbal IQ of low normal range the syntax of Finnish within one school year. In the beginning these boys showed no signs of syntactic knowledge. During the lessons they were shown a visible relationship and its appropriate syntactic marker. The method is essentially the same as the one used by mothers with their infants: “Put it on the table, not under the table”. In this case it only had to be done in writing. At the end of the school year the boys were as proficient as the two best girls in their class. Six years later they proved to be the most proficient their teacher had ever met. Syntax proved to be a skill that could be taught as any other skill. There is nothing innate to it.
CONCLUSION
This post has dealt with one aspect of relativism and universalism – the language relativity and universality. The conclusions drawn from the evidence are that language relativity exists and it must be studied by psychological means. Linguistics does not offer means capable of explaining the psychological reality of language.
Whorf and Sapir were essentially correct in their arguments. It was only that their methods of research were not suited to explain the psychological aspects of language use and the way reality is structured by different languages. It is argued that those in whose opinion all people think in the same way are in error. The search for universal grammar is most probably in vain.
The arguments put forth in this paper may have us to interpret the opinion of Fuller (1991, p. H) in a new light: “….the illusion of epistemic agreement is maintained by a failure to detect real differences that emerge in the process of knowledge transmission. Thus, if it can be shown that the linguistic means at our disposal to transmit truths over time and place is less than reliable, then whatever invariance we seem to find in knowledge systems accepted across socio-historical contexts is unlikely to be due to the invariant nature of the truths transmitted, but rather to cognitive mechanisms – both biases and limitations – that mask the differences in interpretations that would have naturally resulted from truths being unreliably transmitted to different times and places.”
REFERENCES
Chomsky, N. (1957). Syntactic Structures. The Hague: Mouton.
Chomsky, N. (1965). Aspects of the theory of Syntax. Cambridge, Mass.: MIT Press.
Chomsky, N. (1972). Language and mind. New York: Harcourt Brace Jovanovich.
de Saussure, F. (1966) Course in general linguistics. New York: McGraw-Hill.
Devitt, M. and Sterelny, K. (1987). Language and Reality. Oxford: Basil Blackwell.
Fodor, J.A. (1975). The Language of Thought. New York: Thomas Y. Crowell.
Fodor, J.A. (1981). Reprentations: Philosophical Essays on the Foundations of Cognitive Science. Cambridge, Mass.: Bradford Books/MIT Press.
Fuller, S. (1991). Is Relativism Obsolete? Science Studies, 4/2, 5-16.
Gipper H. (1976). Is there a Linguistic Relativity Principle? In: Rik Pinxten (ed.) Universalism and Relativism in Language and Thought. Proceedings of a Colloquium on the Sapir-Whorf Hypothesis. The Hague-Paris: Mouton.
Ihamuotila, P. (1979). Mietteitä suomalaisesta tv-kuvailmaisusta ulkomaanmyynnin kannalta. Helsinki: Oy Yleisradio Ab, unpublished manuscript.
Herriot, P. (1970). An Introduction to the Psychology of Language. London: Methuen.
Kamppinen, M. and Revonsuo, A. (1991). Relativism and Cognitivism. Science Studies, 4/2, 35-43.
Krogh, T. (1991). Relativism and Cross-cultural Communication. Science Studies, 4/2, 25-33.
Matlin, W.M. (1989). Cognition. Chicago: Holt, Rinehart and Winston.
Miller, G.A. (1962). Psychology: The Science of Mental Life. New York: Harper & Row.
Niiniluoto, I. (1991). What’s Wrong with Relativism? Science Studies, 4/2, 17-24.
Penn, J.M. (1972). Linguistic Relativity versus Innate Ideas. The origins of the Sapir- Whorf hypothesis in German Thought. The Hague-Paris: Mouton. (Janua linguarium, Series minor, 120).
Robins, R.H. (1976). The Current Relevance of the Sapir-Whorf Hypothesis. In: Rik Pinxten (ed.) Universalism and Relativism in Language and Thought. Proceedings of a Colloquium on the Sapir-Whorf Hypothesis. The Hague-Paris: Mouton.
Sapir, E. (1929). The status of linguistics as a science. Language, 5: 207-214
Sapir, E. (1929). Selected Writings in Language, Culture and Personality. Edited by D.G. Mandelbaum. Berkeley: University of California Press.
Strømnes, F.J. (1973). A semiotic theory of imagery processes with experiments on an Indo- European and a Ural-Altaic language: Do speakers of different languages experience different cognitive worlds? Scandinavian Journal of Psychology, 14, 291-304.
Strømnes, F.J. (1974a). Memory models and language comprehension. Scandinavian Journal of Psychology, 15, 26-32.
Strømnes, F.J. (1974b). To be is not always to be. The hypothesis of cognitive universality in the light of studies on elliptic language behaviour. Scandinavian Journal of Psychology, 15, 89-98.
Strømnes, F.J. (1974c). No universality of cognitive structures? Two experiments with almost perfect one-trial learning of translatable operators in a Ural-Altaic and an Indo-European language. Scandinavian Journal of Psychology, 15, 300-309.
Strømnes, F.J. (2006) The Fall of the Word and the Rise of the Mental Model. A Reinterpretation of the Recent Research on Spatial Cognition and Language. Frankfurt am Mein. Peter Lang.
Strømnes, F.J. and Iivonen, L. (1985). The teaching of the syntax of written language to deaf children knowing no syntax. Human Learning, 4, 251-265.
Strømnes, F.J., Johansson, A. and Hiltunen, E. (1982). The externalised image. A study showing differences correlating with language structure between pictorial structure in Ural-Altaic and Indo-European filmed versions of the same plays. Helsinki: The Finnish Broadcasting Corporation, Report No. 21/1982.
Whorf, B.L. (1956). Language, Thought, and Reality. Selected writings of Benjamin Lee Whorf. Edited and introduced by J.B. Carrol. Cambridge, Mass.: MIT Press.